MG-PDB & MGBench: Benchmarking co-folding methods for molecular glue ternary structure prediction

Molecular glues (MGs) are a novel therapeutic modality that create or stabilize protein-protein interactions. To benchmark computational prediction of these systems, we present MG-PDB & MGBench, comprehensive datasets of MG ternary complex structures. This work evaluates several state-of-the-art co-folding methods: AlphaFold 3, Boltz-1, Chai-1, Protenix, and RoseTTAFold All-Atom.分子胶（Molecular glues, MGs）是一种新型治疗手段，可构建或稳定蛋白质-蛋白质相互作用。为对这类系统的计算预测开展基准测试，我们发布了MG-PDB与MGBench这两类涵盖分子胶三元复合物结构的综合性数据集。本研究对多款当前最先进的共折叠方法进行了评估，包括AlphaFold 3、Boltz-1、Chai-1、Protenix以及RoseTTAFold All-Atom。